On 27 Apr 2023, at 3:27, Chris Mi wrote:
>>> recirc_id(0),in_port(1),eth(src=06:9c:de:63:c9:40,dst=01:00:5e:00:00:fb),eth_type(0x0800),ipv4(tos=0/0x3,frag=no), >>> packets:13, bytes:1883, used:5.700s, >>> actions:userspace(pid=4294967295,sFlow(vid=0,pcp=0,output=20360),actions),set(tunnel(dst=172.31.1.1,ttl=64,tp_dst=4789,flags(0))),4 >>> recirc_id(0),in_port(1),eth(src=06:9c:de:63:c9:40,dst=1e:51:4f:cd:c3:0e),eth_type(0x0800),ipv4(tos=0/0x3,frag=no), >>> packets:999, bytes:97902, used:0.010s, >>> actions:userspace(pid=4294967295,sFlow(vid=0,pcp=0,output=20360),actions),set(tunnel(dst=172.31.1.1,ttl=64,tp_dst=4789,flags(0))),4 >>> recirc_id(0),tunnel(tun_id=0x0,src=172.31.1.1,dst=172.31.1.100,tp_dst=4789,flags(+key)),in_port(4),eth(src=1e:51:4f:cd:c3:0e,dst=06:9c:de:63:c9:40),eth_type(0x0800),ipv4(frag=no), >>> packets:999, bytes:83916, used:0.010s, >>> actions:userspace(pid=4294967295,sFlow(vid=0,pcp=0,output=20356),actions),1 >>> >>> I did notice a little difference between TC and Kernel, below is the kernel >>> and it has flags as (df) where tc does not have this. Is this a bug in tc? > Forgot to address this comment. I found this comment: > > /* XXX: This is wrong! We're ignoring DF and CSUM flags configuration > * requested by the user. However, TC for now has no way to pass > * these flags in a flower key and their masks are set by default, > * meaning tunnel offloading will not work at all if not cleared. > * Keeping incorrect behavior for now. */ > tnl_mask->flags &= ~(FLOW_TNL_F_DONT_FRAGMENT | FLOW_TNL_F_CSUM); > > Maybe it's a known issue. Yes, this looks like the problem. I guess we should add a note to the tc-offload.rst document’s ‘Known TC flow offload limitations’ section for both the general tunnel offload and sflow section. This concludes my comments to your comments on my comments ;) And sorry for the late response but I was on PTO. Cheers, Eelco _______________________________________________ dev mailing list [email protected] https://mail.openvswitch.org/mailman/listinfo/ovs-dev
